Benchmark it with some real work.
Marshalling to native is expensive when doing micro benches with no work
but not compared to going over the wire or messages that do real work ,
Serialization cost will normally be higher than marshalling to native . If
your doing inter process via shared memory
:21 AM, Bennie Kloosteman bkloo...@gmail.comwrote:
Benchmark it with some real work.
Marshalling to native is expensive when doing micro benches with no work
but not compared to going over the wire or messages that do real work ,
Serialization cost will normally be higher than marshalling
to copy from and to the kernel
, and may even employ unsafe / JNI code to self manage some buffers so
you dont need to pin the GC .. .
Ben
On Fri, Oct 11, 2013 at 12:46 PM, Tony Arcieri basc...@gmail.com wrote:
On Thu, Oct 10, 2013 at 8:56 PM, Bennie Kloosteman bkloo...@gmail.comwrote
If you run it without a network stack zeromq can be very fast so its not
so much the protocol. You can see this in the tests but note in those cases
zeromq is using C++. For lots of small messages 100 bytes zeroMq does
need a lot more cpu than tcp but that is not a common scenario.
There is
Correct, it's only for source code and config files and perhaps
sending by email.
For email you would always use Base64 , there is no efficiency gain due to
the header costs and every system uses it.
For source and config , it would only apply to older languages , most
modern languages have all
The other question is do you want persistant messaging ...To which i would
say no ( in 95% of cases) ...good in theory , crap in the field and
builds the expectation that things just work and when things go pear shape
and it doesnt just work you dont have the systems that deal with
failure.
I wouldn't like to build an actor model directly on ZeroMQ as you can do
better for in memory transport than ZeroMQ and you may want to place some
actors on the same server .. as the implementation for an abstract Actor
framework is fine just don't tie it to the implementation.
Regarding threads
1. Each thread has one socket. You hash the unique name of the actor and
assign it accordingly to a thread.
advantages: no contention.
disadvantages: it cannot work if actors have different processing needs.
You rely on the randomization of the hash function to balance the work.
this is pretty
On Wed, Nov 28, 2012 at 11:19 AM, Pieter Hintjens p...@imatix.com wrote:
I did quite a lot of experimentation with UDP for 0MQ patterns a year
or two ago. The results weren't impressive, performance wise.
Semantically, it maps quite nicely, if you're OK with short messages
and a high rate of
This is a fairly standard bus pattern , though personally i
HATE persistent queues with a passion , they kill all performance , you
have to manage the infrastructure and the possibility of poison messages -
for most business cases there are better ways to guarantee delivery eg send
update events
Ethernet contention ? It is broadcasting
On Tue, Oct 30, 2012 at 3:19 AM, Rohan Bedarkar r_bedar...@yahoo.comwrote:
Steve,
I found something interesting. ZMQ roundtrip over epgm drops suddenly at
one point.. Thoughts?
..
...
88: 143us
89: 144us
90: 145us
91: 146us
92:
On Oct 29, 2012, at 9:14 PM, Bennie Kloosteman bkloo...@gmail.com wrote:
Ethernet contention ? It is broadcasting
On Tue, Oct 30, 2012 at 3:19 AM, Rohan Bedarkar r_bedar...@yahoo.comwrote:
Steve,
I found something interesting. ZMQ roundtrip over epgm drops suddenly at
one point
Why not create the context and socket after you fork ?
Ben
On Tue, Oct 9, 2012 at 2:57 PM, John Khvatov iva...@gmail.com wrote:
Hello all!
I'd like to get some feedback about my report:
https://zeromq.jira.com/browse/LIBZMQ-441
My be there is any workarounds about this problem?
--
WBR,
Doesnt the kernel send an EPIPE if you do a read on a socket where the
other end has sent a TCP RST , lots of routers send this so you need to
handle it.
On Thu, Sep 20, 2012 at 12:22 AM, Yi Ding yi.s.d...@gmail.com wrote:
On Wed, Sep 19, 2012 at 10:49 AM, Chuck Remes li...@chuckremes.com
If you wait for responses you have to wait on latency which can get much
worse eg international 250ms your limit is 4 messages per second. You
cannot do better
That why people build asynch systems...and why Doom/Quake use to spam UDP
position updates - without waiting for an answer.
Ben
On
Unix or windows server
Ben
On Thu, Sep 6, 2012 at 6:02 PM, HP010170 hp010...@gmail.com wrote:
Zhong:
If you can provide the source code, it will be easier for the various
0mq experts on the list to assist with your problem.
HP
On 06/09/2012 06:29, Zhong Weilin wrote:
Hi,
zeromq
Ahh thanks, looks like the anti add bar in brain needs adjustment
Ben
On Fri, Sep 7, 2012 at 10:44 AM, Pieter Hintjens p...@imatix.com wrote:
On Fri, Sep 7, 2012 at 10:41 AM, Bennie Kloosteman bkloo...@gmail.com
wrote:
I see the 15/ZMTP spec is covered by the GPL licence
On Thu, Aug 30, 2012 at 11:56 PM, Julie Anderson
julie.anderson...@gmail.com wrote:
As I said in the text you quoted: I will try to come up with a simple
version to do the same thing.
But Stuart did that for me in C. My thanks to him.
I am not complaining about anything... Just trying to
2) Do people agree that 11 microseconds are just too much?
Nope once you go cross machine that 11 micro seconds become irrelevant .
The fastest exchange im aware of for frequent trading is 80 micro seconds
(+ transport costs) best case , so who are you talking to and if your not
doing
On Thu, Aug 30, 2012 at 10:35 AM, Julie Anderson
julie.anderson...@gmail.com wrote:
See my comments below:
They appear to
be single threaded synchronous tests which seems very unlike the kinds
of applications being discussed (esp. if you're using NIO). More
realistic is a network
What error...is it a unicode error since your not using wchar eg no match ?
Ben
On Mon, Aug 20, 2012 at 3:55 PM, Alexander Voron
voron.alexan...@gmail.comwrote:
I've tried to program sending messages from one server to multiple
clients. I have to use C# on client side, C++ on server side. I
:10, Bennie Kloosteman bkloo...@gmail.com wrote:
What error...is it a unicode error since your not using wchar eg no
match ?
Ben
On Mon, Aug 20, 2012 at 3:55 PM, Alexander Voron
voron.alexan...@gmail.com
wrote:
I've tried to program sending messages from one server to multiple
Do you need messages replayed from the database or just stored ?
If replayed to me its sounds like a queue like MSMQ ( or the IBM one) is a
better fit as it does everything you want , personally i hate queues , the
performance especially for persitant ones is bad and the superficial
simplicty
Just because the socket is guaranteed does not guarantee the receiving app
/ service has received it... Only your own app can do that with some sort
of ack.
Ben
On Wed, Aug 8, 2012 at 6:16 AM, Danil Gazizov danil.gazi...@gmail.comwrote:
Soon.. I can answer to myself - yes, it does, like socket
Like most systems ZeroMQ will just send the message to the transport when
the transport has received the message ( eg TCP/IP )it is removed from
ZeroMQ. There are no guarantees it has even been sent , yet alone that the
transport layer on the other side has the message. This is pretty standard
On Sat, Jul 7, 2012 at 8:16 AM, Mark Farnan mark.far...@petrolink.comwrote:
Hi Ben,
** **
HttpS (more specifically SSL) is mandatory requirement in the Industry,
just forget trying to do anything without it that involves the Internet.
I know but you said minimal authentication , if
Hi Mark,
Web socket uses tcp underneath so getting through the firewall is no issue
just use port 80 , and in fact web sockets ( and things that use port 80 )
have issues with some ISP html proxy servers and load balancers and it will
take time for these products to mature to handle web sockets.
1. Ensure the DLL has permissions to run on your local machines ( right
click properties) . This is not just for .net but exes also.
2. Make sure you are running in full trust .
Regards,
Ben
On Wed, May 2, 2012 at 1:18 PM, Thomas Fee angry...@hotmail.com wrote:
Hey Alex, sounds great but no
28 matches
Mail list logo